RESEARCH PUBLICATION No. 57

Sea Surface Temperatures on the Great Barrier Reef: a Contribution to the Study of Coral Bleaching

JMLough Australian ]nstitute of Marine Science

cRc--).., . AUSTRALIAN INSTITUTf ...... Of MARINE SCIENCE A REPORT TO THE GREAT BARRIER REEF MARINE PARK AUTHORITY © Great Barrier Reef Marine Park Authority, Australian Institute of Marine Science 1999 ISSN 1037-1508 ISBN 0 642 23069 2

-nuswork is oopyrighl Apart from any use as permitted under the Copyright AcI1968, no part may be reproduced by any process without prior written permission from the Grut Barrier Reef Marine Park Authority and the Australian Institute of Marine Science. Requests and inquiries cof'll:eming reproduction and rights should be: addressed to the Director, Wornuolion Support Croup, Great Barrier Reef Marine Park Authority, PO Box 1379, TownsvilleQld 4810.

The opinions expressed in this docwnent are not necessarily those ofthe Greal Barrier Reef Marine Park Authority. Arouacy in c.alculations. figures, tables, names, quotations, refel"!rlCeS ~c. is the romplete responsibility of the author.

National Library of Australia Cataloguing-in-Publication data:

Lough,]. M. Sea surface temperatures on the Great Barrier Reef: a contribution to the study ofcoral bleaching.

Bibliography. ISBN 0 64223069 2.

1. Ocean temperature - Queensland - Great Barrier Reef. 2. Corals - Queensland - Great Barrier Reef - Effect of temperature on. 3. Coral reef ecology - Australia - Great Barrier Reel (Qld.l - Effect 01 temperature on. 4. Great Barrier Reel (Qld.l. I. Great Barrier Reef Marine Park Authority (Australial. n. TItle. (Series, Research publication (Great Barrier Reel Marine Park Authority (Australia)l ; no. 57).

551.460109943

GREAT BARRIER REEF MARINE PARK AuTHoRrrY PO Box 1379 Townsville Qld 4810 Telephone (07) 4750 0700 CONTENTS

LIST OF FIGURES iii LIST OF TABLES iv SUMMARy 1 INTRODUCTION 2 DATA....•...... 2 GBRMPA in situ SST data , 2 IGOSS-NMC SST data 2 COSTA SST data 3 RESULTS 5 Comparisons ofdata sets 5 GBRMPA data loggers and IOOSS blended satellite 5 GOSTA "ships ofopportunity" and IGOSS blended satellite 5 How unusual was the 1997·1998 summer season? 8 Evidence from GBRMPA SST data 8 Evidence from IGOSS SST data 13 The historical perspective from the GOSTA data set, 1903-1"4 15 Average conditions IS Variations associated with ENSO and anti-ENSO events 17 Was 1997-98 a "typical" ENSO for the GBR? 19 Are summer seasons becoming more extreme? 19 Is the SST climate ofthe GBR changing? 23 Evidence from other climatic indices 27 CONCLUSIONS 27 ACKNOWLEDGEMENTS 30 REFERENCES 30

LIST OF FIGURES Figure 1. Location map 3 Figure 2. Weekly SSTs for 5 GBRMPA loggers and closest IGaSS 1° square and difference in SST, (GBRMPA-IGOSS) 6 Figure 3. Comparison of monthly SSTs for Geoffrey Bay flat GBRMPA and IGOSS 7 Figure 4. Annual average

iii Figure 20. Annual average SST anomalies, 1903-1994 26 Figure 21. Linear trend and percent variance explained for seasonal an<;l annual SSTs 28 Figure 22. Annual anomalies of SSTs and other climatic indices 29

LIST OF TABLES Table 1. Summary of daily SST statistics from 8 GBRMPA data loggers 4 Table 2. Number of days and number of degree days daily average, daily maximum and daily minimum SSTs exceed summer season "threshold" for 8 GBMRPA data loggers 14 Table 3. Return periods for maximum monthly mean SST;::: Isd of 1904-1994 mean 25 Table 4. Average maximum monthly mean SST and 10 wannest years 25 Table 5. Correlation between annual SST variations on the northeast Australian coast and other regional climatic indices 27

iv SUMMARY

The extensive coral bleaching event on the Great Barrier Reef(GBR) in early 1998 highlighted the possible causative role ofelevated sea surface temperatw'eS (SSTs) and links with El Nino-Southern Oscillation (ENSO) events. Three questions related to the 1998 event and SSTs offthe northeast Australian coast are addressed in this report:

+ How unusual were SSTs in 19981 • Was this a typical ENSO signal on the GBR? • Is the SST climate ofthe GBR changing?

Three data sets are used to examine SST variations offnortheast Australia: in situ data loggers, blended satellite data and "ships ofopportunity" data.

Agreement between absolute in situ SST data logger and blended satellite observations varies with location and season. Relative temporal variations in the two data sets show a significant level of agreement. Agreement between the two data sets increases with averaging over longer time periods (eg monthly is better than weekly). Agreement between the blended satellite and "ships of opportunity" data is reasonably good. Systematic biases are, however, evident, and are possibly related to common data used to "tune" the satellite data and seasonal and latitudinal variations in cloud amount. Overall, however, the level ofagreement between the three different SST data sets is surprisingly good.

Evidence from the daily SST data loggers located on reefs suggests that 1998 was unusual both in the number ofdays SSTs exceeded a particular threshold and also in the intensity ofthe SST anomalies above the threshold (ie number ofdegree days). Evidence from the blended satellite data indicates that the maximum monthly mean SSTs in early 1998 were the most extreme during the period 1982 to 1998. The next warmest years were 1987 and 1982. Evidence from the "ships of opportunity" data places 1987 as the most extreme year within the period 1903 to 1994. This, combined with the satellite data, suggests that 1998 was the most extreme year in the past 95 years on the GBR.

During a typical ENSO event SSTs on the GBR tend to be cooler than nonnal in the winter season and wanner than Donnal in the following summer. SST anomalies on the GBR in 1997·98 were qualitatively similar to typical ENSO events.

Annual and seasonal average and maximum monthly mean SSTs have increased significantly off the northeast Australian coast from 1903 to 1994. These warming trends are evident along the GBR but are greatest to the south ofthe GBR. These variations match warming over the same time period in Queensland average and minimum temperatures and the Southern Hemisphere combined land and marine temperatures.

It is concluded that:

• SSTs on the GBR in early 1998 were the warmest in the past 95 years ofinstrumental records. • SST anomalies on the GBR in 1997-98 were typical ofENSO events. • SSTs on the GBR have significantly warmed over the present centwy.

1 INTRODUCTION

The extensive coral bleaching event on the Great Barrier Reef(GBR) in early 1998 focussed attention on the role that unusually high sea surface temperatures (SSTs) might play in triggering coral bleaching. 1997~ 1998 also witnessed significant coral bleaching events at many reef sites around the world (Wilkinson" 1998; Berkelmans and Oliver, in press). A link was made between unusually high summer SSTs and the 1997-99 EI NinirSoutbem Oscillation (ENSO) event (Wilkinson et al., in press). As a contribution to the study ofcoral bleaching events on the GBR, this repon addresses the following questions:

1. How unusual were the SST anomalies of 1997·19981 2. What is the average SST signal on the GBR associated with ENSO and anti-ENSO events and was the signal observed during 1997-98 consistent with this? 3. Is there evidence that the SST "climate" ofthe GBR region is cbanging? ie is the frequency of conditions that might be associated with coral bleaching increasing?

These questions are addressed using three readily available SST data sets:

• Great Barrier ReefMarine Pari< Authority (GBRMPA) in situ SST logger data recorded half­ hourly for various sites along the GBR from the mid-1990s. • Integrated Global Ocean Services System (IGOSS)-NMC blended satellite data available weekly and monthly from November 1981 to date for IO-latjwde by longitude square resolution • Global Ocean 8wface Temperature Atlas Plus (GOSTAplus) "ships ofopportunity" data available monthly from 1903 to 1994 for 1°-latitude by longitude square resolution.

Each of these data sets has its advantages; the GBRMPA data logge", are recording SST at high temporal resolution at reef sites, the IG088 blended satellite data is updated rapidly and provides a 1llJ'8C-sca1e, global perspective and the GOSTA "ships ofopportunity" data provide an historical perspective extending back to the start ofthe 20tb century.

Comparisons are first made between the different SST data sets: GBRMPA and IGOSS, lGOSS and GOSTA. Analyses are then presented for to-latitude bands along the northeast Australian coast from 10.5-29.5°8. The latitudinal bands from 10.5-24.5°8 encompass the GBR. The average SST "signa'" associated with ENSO and anti-ENSO events is detennined. Variations in summer warm season extremes and annual and seasonal average SSTs are then examined for evidence of significant trends from 1903 to 1994. Comparisons are then made with other regional climatic indices over the present century.

DATA

GBRMPA in situ SST data The Great Barrier Reef Marine Park Authority (GBRMPA) started a long-term SST monitoring program on the GBR in the mid-l990s. Data are obtained from in situ data loggers at various reef sites. The loggers record SSTs every half-hour and the data are downloaded every 6-12 months (see www.gbnnpa.gov.aulseatempl). Daily average, daily maximum and daily minimum SSTs were obtained for eight sites (Figure I; Table 1; Ray Berkelmans, GBRMPA, pers. comm. October 1998). The choice ofthese siies was based on 1) the record being available through early 1998, and 2} the record being continuous over more than two years. For comparison with the IGOSS data, the daily SSTs were averaged over the same weeks (Sunday to Satw-day) as the blended satellite data.

IG08S-NMC SST data Weekly and monthly SSTs were obtained for to-latitude by longitude squares in the vicinity ofthe GBR from the lGOSS-NMC data set (see http://ingrid.ldeo.columbia.eduD. These data are

2 140"£ i 15O"E -_..:::c::.."1j=r-p~;:r---i ,oos Pacific . Oeean "';.;:;•.~.;l-/Coconu, Beach Reef \ Peloroua Is eatueBayand Pio....rBay

cal""'::g~~'~~(O~"'~heUS• "'_Bay....'s) Townsvtll.. ("'gnetlc Is' 2O"S

Rocklwnpton •• Hero Is

Figure 1. Location map showing l°~latitude oceanic "boxes" for GOSTA and IGOSS SST data and sites ofGBRMPA SST loggers used in this report. available from November 1981 to date (the data are updated by about the second week ofeach month). These data are SST fields blended from ship, buoy and bias-eorrected satellite data based on optimum interpolation (see Reynolds and Smith, 1994). For the period from 1981-1989, the in situ data were taken from the Comprehensive Ocean Atmosphere Data Set (COADS; based primarily on historical reports from "ships ofopportunity". see Woodruffe/ a/., 1993). For the period from 1990 to present the in situ data are from radio messages on the Global Telecommunication System.

Comparisons between the IGOSS and GBRMPA data were based on the 1° square closest to the data logger site. For the comparisons with the GOSTA data and subsequent analyses, the lGOSS 10 square data were averaged for 1°·latitude by 2-4° longitude boxes along the northeast Australian coast from 1O.5'S to 29.S'S (see Figure I). It should be noted that both the IGOSS and GOSTAPlus 1°SST data contain some element ofspatial smoothing so that adjacent squares will not be independent.

GOSTA SST data Monthly SSTs were obtained for 1'·latitude by longitude squares from the GOSTAPlus data set (Version GIST 2.2) produced by the United Kingdom Meteorological Office in collaboration with the Massachusetts Institute ofTechnology and the Physical Oceanography Distributed Active Archive Center (data freely available on CD-ROM on application to the UKMO). Monthly data are available from 1903 through 1994. These data are averages ofSST measurements by month and square made by "ships ofopportunity", supplemented in recent years by Automatic Weather Stations and buoys (see Bottomley etaI., 1990). Instrumental SST measurements from "ships of opportunity" are affected by changes in measurement procedures over the 20111 century (basically a change from uninsulated canvas or metal buckets to either insulated buckets or engine4 intake or hull-sensor thermometers). The latter can record SSTs warmer than insulated bucket measurements resulting in a spurious worJdwide increase in SSTs over the present century. The GOSTAplus SST data used here have been corrected for these measurement (and other) sources oferror (see Folland and Parker, 1995). Monthly SST data were averaged for the same 20 boxes as the IGOSS data (see Figure 1).

3 Table 1. Summary ofdaily SST statistics (OC) from 8 GBRMPA data loggers for the summer (October~March) 6 months ofthe year.

Site Start date End Date Ndays Daily Average Daily Maximum Daily Minimum

Coconut Beach 20 Ju\ 6 Sep Mean 28.3 29.0 27.7 ReefFlat 1996 \998 364 Maximum 3\.8 32.7 3\.2 Date ofmaximum \4 Feb 1998 \4 Feb \998 \5 Feb \998 Mynnidon Reef 14 Apr 2 May Mesn 27.6 28.\ 27.2 Flat \995 \998 547 Max.imum 29.9 30.7 29.4 Date ofmaximum 20 Feb \998 19 Feb \998 \6 Feb \998 Petorus Is 6 Nov 26 Apr Mean 27.9 28.9 27.\ ReefFlat 1993 \998 875 Maximum 3 \.3 32.6 30.7 Date ofmaximum 30 Jan \998 28 Jan 1998 30 Jan \998 Pelorus Is 3Jan 30 Apr Mesn 27.8 28.1 27.5 ReefSlope \995 \998 635 Maximum 30.5 3\.7 30.3 Date ofmaximum \5 Feb \998 12 Feb 1998 15 Feb 1998 Cattle Bay 16Jul 30 May Meso 27.7 28.8 26.9 (Orpheus Is) \994 \998 729 Maximum 3\.5 32.6 30.9 Reef Flat Date ofmaximum 3 Feb \998 8 Feb \998 2 Feb \998 Pioneer Bay 17 lui 30 Apr Mean 27.8 28.\ 27.5 (Orpheus Is) 1994 1998 729 Maximum 30.6 3 \.8 30.4 ReefSloDe Date of maximum 19 Feb 1998 I Feb 1998 18 & 19 Feb 1998 Geoffrey Bay 18 Dec 2 Mar Mean 28.7 30.3 27.3 (MagDetle Is) 1992 1998 1,015 Maximum 32.3 34.7 3\.7 ReefFlat Date ofmaximum 4 Feb \998 8 Feh \998 5 Feb 1998 Reroofs 24 Nov 16Jun Mean 26.0 28.6 24.0 Reef Flat \995 \998 493 Maximum 30.2 35.4 28.6 Date ofmaximum 22 Feb 1998 23 Feb 1998 22 Feb /998

4 RESULTS

Comparisons ofdata sets GBRMPA data loggers andIGOSS blendedsotellite The agreement between absolute weekly SSTs from GBRMPA data loggers and IGQSS varied with location (Figure 2). At Coconut Beach flat and Geoffrey Bay flat, for example. the IGOSS data appeared to consistently record lower summer maxima and higher winter minima, ie dampen the annual cycle. At Pelorus Island flat, the difference between the two data sets was greater for the winter minima than for the summer maxima. At Heron Island, in the southern GBR., the IGOSS appeared to match the summer maxima but record higher SSTs than the data loggers during the remainder ofthe year. At the Myrmidon Reefflat site (located offshore at the shelfedge) there appeared to be remarkable agreement between the in situ data logger and the blended satellite data.

The correlation coefficients between the pairs ofweekly SSTs were all highly significant but will obviously be inflated by the presence ofthe annual cycle in both series. More infonnative about the level ofagreement ofat least the temporal variations ofthe two series (rather than absolute values) were the correlations between the week-to-week changes in SSTs. These were all significant at the 5% level: 0.35 (n=I09) for Coconut Beach flat, 0.54 (n=23I) for Peloros Island flat, 0.61 (n=157) for Mynnidon Reeffla~ 0.45 (n=269) for Geoffrey Bay flat and 0.54 (n~IJ 1) for Heron Island flat. (N.B. for comparison, the correlation between weck-to-weck SSTs at Pelorus Island slope and Pelorus Island flat was 0.85, n= 171).

Thus, the temporal variations in SSTs measured in the two totally independent data sets were significantly correlated. although the strength ofthe relationship varied with location (ie agreement was better at some sites than others).

Comparisons ofmonthly averages for Geoffrey Bay flat (the longest available GBRMPA logger record) with monthly IGOSS data illustrate a number offeatures. First, (Figure 3a and 3b) the dampened annual cycle ofthe IGOSS data compared with the reefflat logger. Second, at the monthly level, a significant level ofagreement (r = 0.71) between the anomalies recorded by the two independent data sets. (ie the level ofagreement was improved with averaging over longer time periods).

GOSTA "ships ofopportunity" and IGOSS blendedsatellite Comparisons between the GOSTA and lGOSS SST data averaged for the latitudinal boxes (see Figure t) were made for the common period, January 1982 through December 1994. Annual average SSTs from the two data sets were in good agreement (Figure 4a) with differences not exceeding O.2°C. The average annual range (between mean annual maximum and minimum SSTs) also showed a very similar spatial distribution along the northeast Australian coast (Figure 4b). Generally, the IGOSS data set showed a slightly dampened annual range (-0.2 to ~.4°C) compared with the GOSTA data set. Both databases clearly showed a local maximum in the annual range from about 21-24°S.

Differences between the two data sets for average monthly values are illustrated for altemate latitude bands in Figure 5. Generally, the IGOSS set recorded cooler monthly mean SSTs to the south ofthe GBR for all months except winter. In the central and northern GBR the lGOSS data set tended to record higher SSTs in winter. A common feature at all latitudes was for the IGOSS to record cooler SSTs in November. Differences in long·tenn monthly means obtained from the two data sets ranged from -ll.5'C to +1l.6"C.

5 3.2 a COCOtlut ~&Il Fill {~1U·SjandIGOSS r "0.13 ,3 GBRMPA .IOOSS

o .j.....-----~__J ., ., • ,3 GBRMPA .IGOSS

".1- - _ • 3Z C ~IoFWI .. flit (-f'.n, andlGOSS r.O.91 ,3 GBRMPA. .IGOSS

" o.J.--. .. ., ., 2O.j.....- ....:..V__~-__ • 32 d G~rey BIy Flatt- ft.1"$) IN taOSS r" 0." , BRMPA. ·lGOSS ~ f " .. I .. .j.....__-'--_.L_~ _ • 31 • HwOll bland Flat (- 2S.S"Sland IGOSS r" 0.•7 , , GBRMPA .IGOS$ " o.j.....-----....,--"-.t,.-,,,,j-- ., " ., ,,+-----~-----'---- • 3OI1vt2 ""... '0112102 ""... n.. """" TIm. """"

Figure 2. Weekly SSTs for 5 GBRMPA loggers and closest IGOSS I' square and difference in SSTs (GBRMPA·lGOSS). Correlation coefficient (r) also given.

6 31 • Monthly m..... SST 1"3-1'98

2t G8fl.MPA .." "" _....,r ...... IGOSS " 0 " J~ J~ .., .., • .. b Monthly...SST e, "'" Mon.. ~ -E ~ ~ .. "E .!! ~ • " 4'!.., Jl .. u .. (: Monthty SST lnomlJl.. O.

_u ,m ,,,. ,... ,...

Figure 3. Comparison ofmonthly SSTs for Geoffrey Bay flat GBRMPAand rGOss for a) 1993-98 monthly means, b) monthly SSTs and c) monthly SST anomalies.

Agreement between monthly anomalies from the two data sets (over the 1982-94 common period) was generally quite high (see Figure 6). The correlation between monthly anomalies ranged from 0.70 to 0.84. (NB not including IGOSS data for 22.5°S which has some sort oferror). Thus, over the common period the two data sets had -50 to 70% variance in common. The major warming and cooling events along the northeast Australian coast were evident in the two data sets though there were some differences. Differences between individual monthly anomalies from the two data sets were as much as ±1.4°C.

The question arises as to whether the differences between the lGOSS and GOSTA SST data sets are random. There are two reasons to suspect that they are not. First, the lGOSS "satellite" data are blended and twled with in situ observations. For the period from 1982 to 1989 these in situ data include "ships ofopportunity" observations from the Comprehensive Ocean Annosphere Data Set (COADS; see Reynolds and Smith, 1994)_ COADS and GOSTA both draw upon the sarne "ships ofopportunity" data. Thus, we would expect the agreement between the IGOSS and GOSTA data sets to be better in the period 1982-1989 compared with the later period, 1990-1994. A second potential source ofdifferences relates to the influence ofcloud cover on satellite estimates ofSSTs. Cloud cover prevents a clear view ofthe sea surface and can, therefore, reduce" confidence in the accuracy ofSST estimates made from satellites. In the region examined here, this should be more ofproblem in the summer months (during the seasonal summer monsoon) compared with the winter

7 ,u p ". ,! -e IAnnuallY.... w n.', E'" •. b "'''.rag_ annulI nno. S w u -l!, ... • W• .. U ...... ' ... 1US ...... -.-20." 2US

Figure 4. Annual average (a) and average annual range (b) for IGOSS and GOSTA SSTs, 1982· 1994, for 1°_latitude bands (see Figuro I) from 10.5-29.5°S. months (which are relatively cloud free) and be greater in the more northerly parts afthe region (where cloud cover is also greatest).

Correlations between the GOSTA and IGOSS monthly SST anomalies were, therefore, examined separately for a) the summer 6 months ofthe year (October to March) and b) the winter 6 months of the year (April to September). The correlations were also examined separately for the periods 1982-1989 and 1990-1994. The results summarized by latitude band are shown in Figure 7. (NB break due to erroneous IGOSS data at 22.5°5). The effect of"tuning" with common data did appear to impact on the magnitude ofthe correlation coefficients. With the exception ofthe far north in summer and the far south in winter, the magnitude ofthe correlations between the two data sets was lower in the more recent period. Second, the effect ofincreased cloud COVeT in the north during summer also appeared to influence the level ofagreement between the two data sets. The differences between the magnitude ofthe summer and winter correlations were most marked north of22°S from 1982-1989 and south of23'S in the later period, 1990-1994.

How uouual was tbe 1997-1998 summer season? Evidencefrom GBRMPA SSTdata The advantage ofthe GBRMPA SST loggers is that they record high-resolution (daily average, daily maximum and daily minimum SSTs) at reef sites, ie where the corals are living. The present disadvantage is the shortness ofthe records. The course ofdaily average SSTs for the 8 data logger sites over the two most recent summer seasons is shown in Figure 8. The 30°C temperature is also marked as this has been suggested to be a threshold for coral bleaching to occur (see, for example, Brown, 1997a) though it appears to be most likely that the threshold for bleaching varies with mean thennal conditions ofthe particular reef site (eg Glynn, 1996; Brown, 1997b; Jones el aI., 1997; Podesta and Glynn, 1997). All sites with the exception ofMynnidon reached this 30°C threshold during late January to mid-February 1998. Also remarkable, given that these sites range from Coconut Beach at ....llfS in the north to Heron Island at -23OS in the south ofthe GBR, was the halt

8 0.' 0.' I 11.5"s f :U.5"1

•., ••• 0.' g 23.~1

0.'

0.0 +"L.I!...... ~llL. •., ••• ••• ••• e15G, 0.' h 25.5"5 •.,

•.• j-ilUI....,.-JllL..-Jl:L,..I!I-'ll.llL 0.0 t.r-...... ,....,....,...-.-~- •., •., ••• •• •.• d 17.•.. Cl.I Irt.S'"S 0.' •.• +"LIlI.,.._....,.-~~ ....._ •• •• •• ••• ••• • 1• .5"1 ••• J21.5"$ ... ..,

.. .fJ!l....I!l.~...... _,..II!._,.J!L •.• t--"e.."r1Ir11...... ,IM.-....IL~ ••• •• ••• •• ,., ...... , $ep Nov Jan lib' May '" .....

Figure 5. Difference in monthly means (1982-94) between IGOSS and GOSTA SSTs for .Iternate 1f>~Iatitude bands. Positive values (light stippling) indicate IGOSS wanner than GOSTA and negative values (dark stippling) iodicate IGOSS cooler than GOSTA.

9 , , .11.5"5,_0.10 1%1.5"5'·O.n

1 • • ., ., ., -, , b 13.5"8 r. 11.74 • g 23.5"$ r. 0.'1

• • ., ., ., ., p , , ,e ... 2$.$"$ r. 0.... -e & sE • • •u '!!, ., ., • ., ., • , '" d '7.5"$ ,_0.70 127.n ,.o.n

• .,

-., ••, 12I.n ,.0.71

., ., 1112 1..... t... 1.. 1110 1"2 , .... 1ta2 '* t'" ,... 1. '"2 , .... Y.u Y-

Figure 6. Monthly SST anomalies (from 1982-1994 mean) for IGOSS (bold) and GOSTA (light) data sets for alternate lO·latitude bands. Correlation coefficient (r) between both series also given.

10 0.' .Summar ,.. :.:....1MZ.fH"__ J .~ '.7 .' . ~ .. '.' . ••• /" ,, e 1HO-1tM , .!!- ••• u " '

e0 ••• 8 ,.• bWinler c ~-" 0 ,.. ~ " -, , :" ~ '.7 , " 0 , " 0 • ~ " ,.. ' ...' ••• ,.• +-----~-----~-- 10.5S iUS 11.$$ 22.$$ 2f.S$ LatItUd. lOutn

Figure 7, Latitudinal distribution ofcorrelations between lGOSS and GOSTA monthly SST anomalies for a) summer 6 months and b)wlnter 6 months and two time periods. in the rise ofthe daily SSTs in early January 1998, This short~tenn cooling followed the exceptional flooding event along the central Queensland coast on January 10rb 1998 associated with a tropica1low and may also be associated with Tropical Cyclone Katrina. Without exception, the maximum observed daily average, daily maximum and daiJy minimum temperatures (Table I) at each ofthe 8 sites occurred in late January or February 1998.

The course ofdaily average, daily maximum and daily minimum SSTs for the longest, continuous record, Geoffrey Bay flat is shown in Figure 9 for the last 6 years. The 30°C temperature is again highlighted, Daily average SSTs at Geoffrey Bay usually reach 30°C each summer (Figure 9a). Similarly, daily maximum SSTs typically reach and exceed for a number ofdays the 30°C threshold (Figure 9b). Daily minimum SSTs have exceeded 30°C in 4 ofthe past 6 summer seasons. This would suggest that 30°C is not a straigbtfOIward threshold for coral bleaching to occur. The second threshold line in Figure 9 is an attempt to develop a threshold relevant to each reef site in the absence of long-term daily SSTs and experimental data. The value represents the average SSTs + Isd for the respective daily average, daily maximum and daily minimum series for the summer 6 months ofthe year (October to March). This was close to 30°C for daily average temperatures, below 30°C for daily minimum SSTs and 2°C above for daily maximum SSTs.

These "thresholds' were computed for the daily average, daily maximum and daily minimum SSTs at each ofthe 8 logger sites. The following two parameters were then computed for each site and SST series: 1) the number ofdays the threshold was exceeded and b) the degree days above the threshold (ie (SST-threshold) summed over the number ofdays the threshold is exceeded). The results are summarized for the 8 sites and available years in Table 2 and pJotted for the longest record, Geoffrey Bay flat, in Figure 10. At most sites the 1997-98 season was characterized by the greatest number ofdays above the respective thresholds and also the greatest number ofdegree days. In some cases, eg Pioneer Bay slope, the number ofdays exceeding the threshold was similar or even greater in 1994-95 compared with 1997-98 (see Table 2). The main difference between

II • cau. Bay Flat (-1U"S1 .. .. f-'==T==-c-!\---- " " " " " ,,1---~--_- --~ .. DMynnldorl Flat (- 1U"'SJ

p " ,e 22 -e ~• -i " ,,1---~--_---_--_ u• ~, " • .,• " " " "J---_-_--~--~ d Ptloruall S~("'I.N) " " " " 22 "jlL------~ "2CI/ll7/'M 20101117 2O!G71t' 2OIll111• " 2OIll71H 20101117 2M171t1 20101'" 2MI71l1 nm...... no.

Figure 8. Daily average SST, for 8 GBRMPA logge" from 20· July 1996 through 1998. 30'C shown as dashed line. these two years at this site was the higher number ofdegree days. Thus, although based on a very short number ofyears, the available daily SST data for reefsites along the GBR indicates that the 1997-98 summer season was unusual both in the number ofdays SSTs were above given thresholds and also in the intensity with which SSTs exceeded the thresholds.

12 :u , Geoff~ Sty ftIt dIIlly ,.,.... SST

" "

" -1---_--_---'_--~-~ " .. "

2O-I--_-_---'-_-~-:.~ "

22

,,-I-_~-_-.L_-_-~ 11112/92 1t112J9S 18t12ItA-.11112115 1111ZIM 1111V17

Figure 9. Geoffrey Bay flat (Maguetic Island) ~ daily average, b) daily maximum and c) daily minimum SSTs, ISlb December 1992 to 2 March 1998. Dashed horizontal line is 30°C and solid borizontalline is respective temperature threshold (see text).

Evidencefrom IGOSS SSTdata The IGOSS data set provides infonnation about SSTs for the most recent 17 austral summer seasons along the northeast Australian coast. Studies such as Podesta and Glynn (1997) have demonstrated that various measures ofwann season SSTs are all inter-related. Here one ofthese parameters, the maximum monthly mean SST observed in eacb summer season. is used to assess how unusual the 1997-98 season was in the vicinity ofthe GBR. Maximum monthly mean 88Ts for each latitude band were extracted for each summer from 1982-83 to 1997-9S. Anomalies from the 17-year mean value are plotted in Figure 11 for alternate 1°-latitude bands,

The 1997-98 summer maxima were the wannest ofthe 17-year record for aU latitude bands except 11.5-13.5°8 and 16.5-17.5°8. For the latter two regions, 1997-98 ranked either the second or third most extreme maxima during the record period. The next most extreme summers were either 1981­ 82 or 1986-87. Averaged over the GBR region (10.5-24.5°8; 88Ts tend to vary in a similar manner along the length ofthe GBR; see Lough. 1994, 1998), 1997-98 was the most extreme year with a

13 Table 2. Number ofdays and number ofdegree days (in brackets) daily average, daily max.imum and daily minimum SSTs ex.ceed summer season "threshold" for 8 GBMRPA data loggers. - indicates no data and· indicates incomplete season.

Threshold 'C 1992·93 1993·94 1994·95 1995-96 1996-97 1997-98 Daily Average Coconut Beacb ReefFlat 29.6 -- -- 13 (4) 35 (40) Myrmidon Reef Flat 28.9 - - - 19 (6) 0(0) 39 (20) Pelorous Islaud ReerFlat 29.3 · 3I (9)' 15 (4) 22(6) 6 (I) 35 (38) Peioroul Island Reef Slope 29.2 ·· - 23(3) 0(0) 33 (25) Cattle Bay (Orpheus Is) Reer Flat 29.2 - - 28 (II) 28 (6) 7 (2) 37 (43) Pioneer Bay (Orpheus Is) ReerSlope 29.2 -- 36 (13) 33 (8) 5 «I) 32 (30) Geoffrey Bay (Magoetic Is) ReefFlat 30.2 17(5)' 26 (22) 20 (12) 15 (3) 7 «1) 54 (49)' Heron Is Reef Flat 27.7 -- · 22112)' I1(4) 47(42) Dally Minimum Coconut Beach Reef Flat 29.1 - - · - /5 (4) 32 (39) Myrmidon Reel Flat 28.5 - · - 22 (8) 2 «I) 38 (19) Pelorous bland Reef Flat 28.5 - 35 (12)' 14 (5) 22 (8) 9 (3) 38 (37) Pelorous bland Reef Slope 28.8 - - - 24 (5) I «I) 17 (17) Cattle Bay (Orpheus Is) ReerFlat 28.5 - · 20 (8) 26 (5) 9 (3) 43 (44) Pioneer Bay (Orpheus Is) Reef Slope 29.0 - - 52(15) 30 (8) 5 «I) 29 (21) Geoffrey Bay (Magoetic Is) ReefFlat 29.0 24 (12)' 30 (23) 20 (12) 10 (3) II (3) 48 (53)' Heron Is ReerFlat 25.9 - · · \2(7)' 15 (6) 49'(46) Daily Maximum Coconut Beach Reef Flat 30.4 - - - - 18 (4) 39 (42) Myrmidon Reef Flat 29.4 - · - 23 (8) 2 «1) 44 (25) Pelorous Island Reef Flat 30.3 - 33 (13)' 16 (4) 26(11) 4 (I) 39 (44) Pelorous hland Reef Slope 29.5 - - · 12 (4) 2 «I) 32 (35) Cattle Bay (Orpheus Is) ReerFlat 30.3 · - 35 (23) 13 (8) 9 (2) 42 (45) Pioneer Bay (Orpheus II) Reef Slope 29.6 - - 32 (12) 25 (9) 3 (1) 34 (37) Geoffrey Bay (Magnetic Is) ReerFlat 32.0 14 (9)' 46 (31) 28 (20) 21 (12) 7 (3) 43 (46)' HerOD Is ReelFlat 30.9 · - - 28 (29)' 14 (13) 39 (62)

14 • GeotTrey Bly Flitdally ....erage SST ..50 .. /lIIdtOMdlYS " " • b Geofl'rey Bay Flat dally llIUh'lllJlIl SST 50 •~ • " .. !• .. ..• " " "• •~ " ~ • 0 - c GeoffNy Bay FlM cially mlllllllum SST .li" 50 ,E .. Z .. " " • '"2·'S ,...... ,....., SumrnrMUon

Figure 10. Number ofdays and number ofdegree days above "threshold" for Geoffrey Bay Flat (Magnetic Is) for a) daily average, b) daily maximum and c) daily minimum SSTs. Note, 1992~93 and 1997-98 are incomplete seasons with only 104 and J53 days ofa possible 182 days, respectively. maximum monthly mean SST anomaly of+O.72°C, followed by +O.53°C in 1986~87 and +0.45°C in 1981-82. Thus, the evidence from the lGOSS data set indicates that the 1997·98 summer season had the wannest SST maxima on the GBR over the period 1982 to 1998.

The infonnation obtained from the IGOSS and GOSTA data sets on SST maxima variations was not, however, in total agreement. Anomalies for the region oftile GBR (10.5-24.5°5) with respect to the common period 1982·1994 (Figure 12) showed some conflicts between the two data sets, eg 1981-82 was comparable to 1986-87 in the IGOSS data set but close to average for the GOSTA data set. The correlation between these two average series was 0.79. When examined by latitude band (not shown) the largest disagreements between the two data sets occurred between 10-15.5°$ with -3540% variance in common. Agreement was better from 16.5-26.5°S with -50-80% variance in common.

The historical perspective from the COSTA data set, 19Q3..1994. Average conditions Average monthly SSTs for alternate 1··latitude bands along the northeast Austra1ian coast are shown in Figure 13. Maximum monthly mean SSTs usually occur in January-February with minimum monthly mean SSTs in July-August. The mean annual cycle varies from a rather broad

15 • 11.5'"1 ....n .. 2t.:s"C f21.S"S ...... 21.1·C

-. _1 b U.$'"5 ....n.. 2t.i·e ,,13-" MNn" 27.~C

-, -. •u ,! t 15S$ ....n.2t.o-t: h 2$.5"$ ....It. 27.o"c -~ •a. i , , -•u ~, • • -. _1 '" d17.11·S ....n.2I...C I 21$$ ....n .. 2I.5"c • 1 ,

_1 -. • tI.n ....n .. ZI.5"C I 21.5"$ MeaIl" :K.O"C 1

_1 _1 .... 1. 111% ...... to .... ".. '''' '''' V••

Figure 11. Maximum monthly mean SST anomalies, 1982-98. from IGOSS data set for alternate 1° ·Iatitude bands. Mean maxima also given. peak. in summer in the north (with maxima occurring for example from December-February) to a sharper peak in the south with the maxima generally occurring in February. The average annual

16 l" ,.• f .a IGOSS e ••• /' !. E S •••

•0 ~~ ~, • -1.11 , , til ... ,- ,- ".. ... '"' ,- V-

Figure 12. Comparison ofGOSTA and IGOSS maximum monthly mean SST anomalies from 1982­ 1994 mean for 10.5-24.5·5. range based on the 1903-1994 "climatology" was typically 4_6°C (see Figure 4b). At any given latitude the difference between the observed maximum monthly mean SST and observed minimum monthly mean SST over the period 1903-1994 was about 3°e larger, ie from 6-9°C (Figure 14). This might, therefore, be a more representative "window" ofthe SST range within which coral reef ecosystems ofthe GBR operate than the average annual range (cfFigure 4b).

Variations associated with ENSO andanti-ENSO events

Monthly SST anomalies (from the respective 1903-1994 monthly means) were calculated for each latitude band. These were then averaged from January to December ofthe following year (24 months) for 23 ENSO events and 23 anti-ENSO events (eg for the 1982-83 ENSO event, average anomalies covered January 1982 through December 1983). This "compositing" technique emphasizes SST anomalies common amongst the different ENSO (and anti-ENSO) events and de­ emphasizes those SST anomalies that are not common. The average monthly SST anomalies were then tested for significant difference from the long-term mean (Mitchell et01., 1966). The resulting patterns are shown for alternate latitude bands for ENSO events (Figure IS) and anti-ENSO events (Figure 16).

There are two components to the "average" SST anomaly signal associated with ENSO events along the northeast Australian coast (see also Lough, 1994). These are cooler than nonnal SSTs in the prior winter that extend to about the southern tip ofthe GBR and warmer than nonnal SSTs in the summer season that extend from 10.S·29S'S. In the far north (eg Figure lSa and )Sb) this anomalous warming extends across the nonnal seasonal SST maximum. Further south (eg Figure 1Sd and 15e) the anomalous wanning occurs after the nonnal SST seasonal maximum ie late summer.

Although there was a tendency for anti-ENSO events to be associated with opposite sign anomalies to ENSO events on the northeast Australian coast, the magnitude, extent and significance ofthe SST anomalies were much less (Figure 16). Thus, significant wanning in the prior late winter-early summer season was largely confined to the northern GBR (Figure 160, 16b and 16<). Significant cooling in the summer season was not widespread or significant.

Figure 17 summarizes the "average" ENSO and anti-ENSO signals over the GBR region (10.5­ 24.5"$).

17 ,. .. f :u.rs ....nto 2$.O"C ,. " " .. J-~_~__~~~~ _ "l-_~-_~_~__~_~~~

" bU,5"S ....n.a.re " g23.I-S ....n. 24A"C ,. .. .." ..

"J-~~__~~ ~

" h :n.5"s "..n" n.'·c .. ,. .. ..

"J-.~~ ~ " J-. ~ _ _

" d 17.5·$ Mean. 2S.'"C " " .. ,. .. ..

.. J-~__~~_~ _ "J-..,...l ~__~_~~

~ 21 .tI.n ...... 25•• " I 2I.rs ....,.22.ft " " " ..

J,' Sop ...... Mo,

Figure 13. Monthly mean SSTs, 1903-1994(± Isd shown as eITOr bars), from GOSTA data set for alternate 10-latitude bands. Mean annual SST also given. Note different temperature scales.

18 "r=~/maximum ;;: " ...... ",...•. 22 t----"·~'_.=---_===_-- '&...... 11I"I 1m U", ...... " .. . " ,,'------~------.:.... '. 10.5$ 1$.5$ 20" 25.1S l.lttltud. _ttl

Figure 14. Observed maximum and minimum monthly mean SSTs, I903~ 1994 for each 10-latitude band from GOSTA data set.

Was /997-98 a "typical" ENSOfor the OBR? No two ENSO events are alike (eg Allan et al., 1996). Each evolves in a slightly different manner. There are, however. features (cfFigures 15 and 17a) that appear characteristic ofsuch events in the vicinity ofthe GBR Here I examine whether the 1997-98 was a typical event for the GBR region.

Observed SST anomalies (from lGOSS) for four reoent ENSO events are compared with the average ENSO "signal" for the GBR region (l005-24o5'S) in Fignre 18. The 1982-83 event clearly followed the average signal with prior winter cooling and summer warming (Figure J8a). The overall correlation between the "average" signal and observed SST anomalies for the 24 months was 0.59. The magnitude ofthe correlation varied spatially (not shown), being greatest (>50% variance in common) north of 16.5°5. The 1986-87 event was not typical, with no evidence ofprior winter cooling and only a slight warming evident in early summer (Figure ISb). The overall correlation was -0.19. The 1991-92 event was also atypical with prior cooling only evident in very early winter and no marked wanning in summer (Figure 18c). The 1997-98 event did, however, show strong similarities with the "average" ENSO signal (figure ISd) with an overaJl correlation (over 18 months) of0.79. Prior winter cooling was weakly evident followed by marked summer warming. Spatially, greatest agreement between the average signal and the observed SST anomalies occurred south of 18.5°5 (>50% variance in common). Thus, the 1997-98 ENSO event was associated with a reasonably ''typical'' SST anomaly signal along the northeast Australian coast.

In contrast to the SST anomalies along the northeast Australian coast, summer rainfall variations in Queensland did not follow the typical ENSO signal in 1997-1998. Northeast Australia is one ofthe regions around the world considered to experience consistent patterns ofrainfall anomalies in association with ENSO and anti-ENSO events (eg Ropelewski and Halpert, 1996). ENSO events are usually associated with a weakened summer monsoon and drier than nonnal conditions with a strengthened monsoon and wetter than nonnal conditions in anti-ENSO years (Lough, 1994). Rainfall in the 1997-1998 summer season was close to average for Queensland as a whole and well above average for certain regions (see Climate Monitoring- Bulletin, Australia, Issue No. 143 and 146, Australian Bureau ofMeteorology).

Are summer seasons becoming more extreme? Superimposed on decadal variability (illustrated by the 10-year Gaussian filter) maximum monthly mean SST anomalies significantly wanned over the 91-year record period, 1904-1994, at all latiwdes from 13.5 to 29.5OS· (Figure 19). Wanning was greatest to the south ofthe GBR (figures

1 NB No allowance has been made for the presence ofautocorrelation in assessing the significance oflinear trends. This may cause the level ofsignificance to be overestimated.

19 0.' 0.' • eNSO11.n " " " " 0.0 4"-rrr-wlnn.....rl'rr"\I-ILL 0.0 m..-.r4rn- " " ••• " ••• " 0.5 b ENSO 1:J.1I-S " 0.' 27 " " "

••• 21 0.5 ,EHSO 1S.rs " 27 " hENSOlSJ"S 27 " 0.0 t'1nnmrTlrnhr.,lLlJLU""-/D}"'-aj 0.0 mr'*"lTlO'Trr!rJl,l1JJ.J..I"I.JJJlJ,IUlJ'--I " " ••• " •• 21 0.5 d ENSO 71.5~ " 0.' " 27 " 0.0 Mnrn-nrn-n-".JLILIllUL...,J'.I!f""J " " •• " .... " 0.5 • ENSO ".n u " JENSO 2t.s-S " " UmnnnnrTlrnlrr-.,lIJILUt"-",J'-'j1-"l 0.0 1mJr""\\"""'1T11,fl1,11Jl.Ll,1I-'-1LI,JlLI'-"! "

u "", " "

Figure 15. Average monthly SST anomalies (bars, scale on left-hand axis) associated with 23 ENSO events and mean annual cycle (line. scale on right-hand axis), for alternate Io~latitude bands from GOSTA data set. Anomalies in bold are significantly different from the long-term mean at the 5% level.

20 0.$ aAn6.£NSO'1n

.. .. ~. " o.s b AnU-£HSO n.n .. 0.' " " " ~. " .. "

•• .j!UUL-'1-.-"-JI,ll.A/-_ " " ~.. " ~. " 0.5 d Antl-EHSO 17.n " .. " " 0.' " " ~.• " 0.' " .. " os

20

Figure 16. Average monthly SST anomalies (bars. scale on left-hand axis) associated with 23 anti­ ENSO events and mean annual cycle (line, scale on right-hand axis). for alternate 10-latitude bands from GOSTA data set. Anomalies in bold are significantly different from the long-term mean at the 5% level.

21 ••• • GSR ENSO " U 21 ••• P e, ~~ " -•e ~.. .. OA " -•E bG8R~ " •U U ~, 27 • .. U1• ~~ "

~A Jo. ..., ... --, ...,., ....' " Figure 17. Average monthly SST anomalies (bars, scale on left-hand axis) for GBR (10.5-24.5'5) associated with a) 23 ENSO and b) 23 anti-ENSO events, from GOSTA data set. Mean annual cycle also shown (line, scale on right-hand axis). Anomalies in bold are significantly different from the long-tenn mean at the 5% level.

19h, 19i, and 19j) where the linear trend accounted for ~ 30% ofthe variance. There has been no significant wanning ofsummer season SST maxima in the far north ofthe GBR (Figure 19a).

This increase in summer season extremes was also demonstrated by the frequency ofyears in 3 sub-­ periods in which the maxima was 2:1 sd above the long-tenn mean (Table 3). In the fllSt 30 years of the record such extremes occurred in 10% or less years along the GSR (10.5-24.5°5), ie about once every 10 years. In the most recent 3I-year period, 1964-1994, such extremes occurred about once every four years. The greater warming trend to the south ofthe GBR was also evident with such extremes occurring about every two years in the most recent period compared with only about once every 30 years in the earlier part ofthis century.

The ten years ofmost extreme monthly mean SST maxima (from 1904 to 1994) for each latitude band are listed in Table 4. SST anomalies in 1987 were the most extreme ofthe 91-year record period throughout most ofthe region. For the GBR (10.5-24.5'5) the years in whicb at least 10 of the IS boxes had SST maxima ~ Isd above the mean were: 1910, 1924, 1935, 1946, 1958, 1963, 1964, 1970, 1972, 1977, 1980, 1986 and 1987. 1980 and 1987 were bolb years ofdocumented coral bleaching on the GBR (Betkelmans and Oliver, in press). Although 1982 was also a documented bleaching year (ibid), it does not appear to be unusually warm in the GOSTA data set but was unusually warm in the lGOSS data set (see Figure 12). The reasons for this difference between the two data sets need to be examined further.

Thus, evidence from the GOSTA data set for the period from 1904-1994 suggests that summer season SST extremes are becoming more frequent over most ofthe GBR but especially to the south. Given that 1987 was the most extreme summeroD the GBR within the period 1904 to 1994, and 1987 was the second most em-eme year after 1998 in IOOSS over the period J982-] 998, J998 was likely to have been the most extreme year during the past 95 years.

22 OA • ttl24Sr.O.s, ,.. ••• U .. ••• ••• ~.. A.' .1.0 A' .1.5 OA U b1~r·~.11

~ ••• •• ••• e0 -e ••• ••• i!. ~.. E ~. • .1.0 -•u i" ~. .f.$ 0 OA c:,",,,2r.O.1. ... • .,• U ••• ••• ••• ••• A.' ~. ·U

~. .1,S

••• dtll7-N,aO.n ..

U ••• ••• ••• ••• ~ A. .. ·1.0

A.' .1.~ ...... Jaft

Is the SST climate ofthe GBR changing? Evidence presented in the preceding section related only to SST maxima during the summer warm season. Here, 1examine whether these changes are part ofan overall change in the SST climate of ofthe GBR. Annual average (January to December) SSTs significantly increased at all latitudes along the northeast Australian coast from 1903 to J994 (Figure 20). The warming appears to have occurred mainly since the late 19505 at all latitudes. The magnitude ofthe warming trend was least in the far north ofthe GBR(eg Figure 20a and 20b) where it accounted for less than 20% ofthe variance and was greatest to the south ofthe GBR (Figure 20h. 20i and 20j) where it accounted for more than 40% ofthe variance.

23 ... f Ma>:l_ t1.S'"'S It". 0.11

·U ·u u 1.& bMulllla".5"S R"_O.01 ,Maxima n.n Iit'. 0.11

·u -1.'... 1.1 cMu,",'15n R"_O.tt IlllUlma 25.n lIt'. O~1 ••..

.,.,... .t.' d IIUlfllal1.n It". 0.01 ••• I Maxlma 21.5"5 It'. 0.35 ••• ••• •• _i.' -1.' U ... .. • MaxIma un lIt'. 0.1' JIiIulIN n.n R',. U.S' ..

·u ,... ,... ,.. '02' ......

Figure 19. Maximum monthly mean SST anomalies. 1904-1994, for alternate 10-latitude bands. Thick line is IO~year Gaussian filter. Dashed line (where drawn) is significant Jinear trend and 2 variance explained by linear trend (R ) value also given.

24 Table 3. Return periods (in years) for maximum monthly mean SST?: lsd of 1904-1994 mean for IO·latitude bands and three sub-periods.

'OD 1904-1933 1934-1963 1964-1994 lOSS 10 6 4 l1SS 10 8 3 12SS 10 6 4 13SS 14 6 4 14SS 10 6 4 15SS 14 4 4 16SS 14 5 3 17SS 14 5 4 18SS 10 6 4 19.5°8 14 6 4 20.5'S 10 6 4 21.5'S 10 6 4 22SS 10 6 4 23.5°8 10 8 3 24SS 30 8 2 25SS 30 8 2 26SS 30 8 2 27SS 30 10 2 28.5"5 14 10 2 29SS 30 10 2

Table 4. Average maximum monthly mean SST (GOSTA, 1904-1994) and 10 warmest years for 1°· latitude bands.

RegioD MeaD RaDk 'C 1 2 3 4 5 6 7 8 9 10 10.5'S 29.2 1990 1940 1929 1970 1944 1983 1987 1973 1913 1941 l1SS 29.3 1940 1929 1983 1987 1970 1944 1990 1973 1913 1946 12.5·S 29.2 1987 1983 1940 1970 1929 1986 1973 1980 1971 1946 13.5°8 29.1 1987 1983 1940 1970 1986 1929 1980 1973 1971 1958 14.5'S 28.9 1987 1983 1970 1986 1940 1973 1958 1964 1971 1980 IS.S'S 28.9 1987 1970 1983 1910 1986 1973 1958 1964 1935 1940 16.5'S 28.8 1987 1910 1970 1958 1973 1935 1964 1986 1971 1%3 17.5'S 28.8 1987 1910 1935 1958 1970 1964 1973 1986 1929 1938 18.5"5 28.5 1987 1910 1935 1958 1970 1%4 1973 1938 1977 1986 19.5'S 282 1987 1910 1935 1958 1970 1973 1%4 1994 1986 1977 20.5"5 27.9 1935 1910 1987 1970 1958 1986 1938 1973 1977 1971 21.5"5 27.8 1935 1910 1987 1938 1970 1986 1958 1973 1977 1980 22.5'5 27.5 1935 1910 1987 1938 1986 1970 1980 1958 1977 1973 2305'S 272 1935 1910 1938 1987 1980 1986 1970 1958 1982 1973 24.5'5 26.8 1987 1982 1980 1938 1986 1935 1970 1958 1973 1985 25.5'S 26.4 1987 1982 1980 1958 1986 1970 1985 1973 1938 1992 26.5'S 26.2 1987 1980 1982 1958 1973 1970 1943 1992 1985 1938 27SS 26.0 1987 1982 1980 1973 1958 1970 1943 1990 1992 1938 28.5'S 25.7 1987 1982 1973 1980 1958 1970 1943 1981 1971 1990 29.5'S 25.5 1987 1973 1982 1958 1970 1980 1919 1981 1943 1971

25 ••• • AnnlMl11.rS ~.o_ ••• f Annual 21.5"S ft-. 0.2• ..OA ..••

~. ~.

~. ••• .. Il'AIlnuIIU.rs R'.o.1. .. II Annual nrs Jt". 0.)1 ..OA ..OA

~ ~ •• •• •U ,e .. c Annllal1s.n It". OJll .. II AM_IH.S"S R". 0..1 -~ OA OA 8- E S .. ••• G U ~, ~. .. • G• ~ en .. ••• .. dAnnuaI17.I". 1It'.OOU ••• IAnn".IU.5"S !t'.IlA7 ••• ••• .. •• ~. •• •• ~. .. • Annuaj ".n til'. UCl ••• JAnnual2U·S R'-U' ••• OA •• ••• ~.• .... •• ~. " ".. ,... ,... ,... ,- ,m ,... ,... ,... .. y- y-

Figure 20. Annual average SST anomalies, 1903-1994. for alternate 1°-latitude bands. Thick line is I()..year Gaussian filter. Dashed line (where drawn) is significant linear trend and variance explained by linear tmld (R') value also given.

26 The wanning trend evident in annual and maximum SSTs was also evident at other times of year (Figure 21). Warming was. however. greatest (up to O.I°C per decade) along much of the GBR in the late summer months (JFM, Figure 21a) and was weakest in early winter months (AMJ. Figure 2lb).

Thus. there appears to be good evidence that SSTs along the GBR have warmed over much of the present century especially in the summer season. Wanning has been greatest in magnitude south of theGBR.

Evidencefrom otherc/iJ1Ullic indices Annual anomalies ofSSTs for the GBR region (10.5-24.5"S) and south of the GBR (25.5-29.5"S) were compared with temperature and rainfall variations in Queensland (Lough, 1997), combined land and sea temperatures for the Southern Hemisphere (Houghton et ai., 1996) and the Tahiti-Darwin index of the Southern Oscillation (SOl). Significant warming trends were evident in Queensland average and minimum (nighttime) temperatures and the Southern Hemisphere temperature index. (Trends in Queensland temperatures in recent decades appear to be the largest in comparison with other parts ofAustralia eg www.bom.gov.aulbmrcJrnrlr/daj/tempanal.htnil. There were no significant trends in the SOl or Queensland rainfall. Annual SST variations on the GBR and to the south were significantly correlated with all three Queensland temperature indices, though the magnitude of the correlations was greatest for minimum and average rather than maximum temperatures (Table 5). The correlations were also significant for the original and residual (after removal ofdecadaJ variability) data Thus. SST variations along the northeast Australian coast were similar to those of the adjacent land area (Figure 22). Weak. but significant, correlations also occurred with Queensland annual rainfall and the SOl for SST variations of the GBR region (10.5­ 24.5"S) but not for SSTs to the south of the GBR. Annual SST variations, especially those to the south of the GBR, all appear to track those ofthe Southern Hemisphere land and marine temperatures. This suggests that the increase in SSTs over the present century may be part of, at least, a hemispheric phenomenon.

Table 5. Correlation between annual SST variations on the northeast Australian coast (GBR and region to south of GBR) and other regional climatic indices. Correlation coefficients in brackets are for the residual series from the lO·year Gaussian filter (see Figure 22). Period is 1903-1994 except for Queensland temperature indices, 1910-1994. Values in bold are significant at 5% leveL

Climatic index SST 10.5·24.5"S SST 25.5-29$S Queensland average temperature 0.47 (0.47) 0.58 (0.57) Queensland maximum temperature 0.28 (0.26) 0.37 (0.41) Queensland minimum temperature 0.52 (0.53) 0.63 (0.57) Queensland rainfall 0.27 (0.27) 0.06 (-0.01) sa land & marine temperature 0.50 (0.16) 0.74 (0.41) Southern Oscillation Index 0.21 0.31 -0.14(-0.02)

CONCLUSIONS

The ideal SST data set to assess the nature and significance oflinkages between SST characteristics and variations ofcoral reef phenomena, such as coral bleaching, would have the following characteristics:

• At least daily temporal resolution to allow assessment of both the duration and intensity of SST extremes above and below particular thresholds. • Be spatially extensive.

27 Figure 21. Linear trend ('C increase from 1903-1994, bar, scale on left-hand axis) and percent variance explained (line, scale on right-hand axis) for seasonal and annual SSTs. Trends arc not significant at 5% level for unfilled bars.

28 1.' • SST 10.24'"$ RI • 0.13 ••• f SOuttMm Hemlsph_lan4 & Mal1roe temperatu,. R'-O.70

.1.0 •• 1.0 IISST2540"SR'"OM " 9 Souttlem Oscillation Inoell. " • ." .1.0 ." U c QIl....lWld -neetemperatu,. R1• 0.15 )0 h Ql.I..nsland ralnfalllnftlt 1.' " ••• -1.0 ... •2.0 ... 1S43 ,.63 11113 2.0 d QlI_lUId maxhnllm temperature " .. V..r 1.' •••

·1.0 •• u • Qu..nal..d mlnlmum tempe..-tu.. ~. 0.24 1.'

•.• -t--ri~ .. ".. "" 1m 1m

Figure 22. Annual anomalies ofa) SSTs IO.S·24.5'S, b) SSTs 2S.S-2905'S, c) Queensland average temperatures, d) Queensland maximum temperatures, e) Queensland minimum temperatures, f) Southern Hemisphere land & marine temperatures, g) Southern Oscillation Index. and h) Queensland rainfall index. Thick line is 10·year Gaussian filter. Dashed line (where drawn) is significant linear trend and variance explained (R2) also given.

29 • Provide measurements at various depths and locations on coral reefs. This would provide information about the absolute temperatures experienced by corals - which will vary in intensity depending on local reefconditions (eg tides, extent offlushing oflagoon water, vertical temperature stratification etc). • Be available in near real time so that "critical" conditions can be quickly identified. • Be of long duration (at least 30 years is needed to detennine "climatological' conditions and variability about the average) so that the extent to which particular SSTs (eg January-February 1998) are unusual can be assessed.

Such an ideal data set is not, as yet, available. Elements ofeach ofthese ideal characteristics are, however, found in the three data sets used in this report. There was a significant (possibly surprising) level ofagreement between the different data sets; ie they seem to be telling parts ofthe same story. Despite their respective weaknesses it was possible, using these data, to answer with reasonable confidence the three questions posed at the start ofthis report:

1. SST anomalies ofthe 1997-1998 summer season may well have been the most extreme in the past 95 years. 2. ENSO events are typically associated with prior winter cooling and summer warming ofSSTs on the GBR. Generally opposite, though less marked, SST anomalies are associated with anti­ 2 ENSO events • SST anomalies on the GBR during the 1997-1998 ENSO event were at least qualitatively similar to the "typical" ENSO pattern fOT the region. 3. SSTs offthe northeast Australian coast have significantly warmed from 1903 to 1994. This is evident in annual and seasonal averages and summer season extremes. The frequency of extreme summer season maximum monthly SST anomalies is substantial1y greater in the most recent 30 years compared with the frrst 30 years ofthis century. The warming trend is greatest to the south ofthe GBR and also matches changes over the same time period in average and minimum temperatures over Queensland and the Southern Hemisphere land and ocean temperatures.

The rise in SSTs ofthe GBR over the present century has increased the frequency ofwann temperature extremes during the summer season and thus, potentially, the risk ofcoral bleaching events. Average SSTs in the vicinity ofcoral reefs are projected to rise 2_3°C above pre-industrial values by the end ofthe 2lSl century (eg Houghton e/ aJ., 1996; Pittock, in press). Ifthe observed SST rise on the GBR is a response to the enhanced greenhouse effect then the frequency ofwann temperature extremes is likely to continue to increase. Such an increase could result in an increasing risk ofcoral bleaching events and a reduction in the time for coral reefs to recover from bleaching events. In the absence ofadaptive changes by coral communities this is also likely to increase reefs' vulnerability to other natural and unnatural stresses (eg Wilkinson and Buddemeier, 1994).

ACKNOWLEDGEMENTS

Thanks to Samroai Tobin for preparing Figure I. I thank. Ray Berkelmans for providing the daily GBRMPA SST data and for his helpful comments and encouragement

REFERENCES

Allan, R., J. Lindesay and D. Parker, 1996. El Nino Sou/hem Oscillation & Clima/ic Variability. CSIRO Publishing, Collingwood, Australia, 40Spp.

2 Current indications are that 199&-1999 will be an anti·ENSO event. It is., therefore, Wllikely that SSTs on the GBR during the 1998-1999 sununer season will be unusually warm (see Figure l7b).

30 Berkelmans, R and J.K. Oliver, in press. Large scale bleaching ofcorals on the Great Barrier Reef. Coral Reefs. Bottomley, M., C.K. Folland, J. Hsiung, R.E. Newell and D.E. Parker, 1990. Global Ocean Surface Tempera/ure A/las "GOSTA ". United Kingdom Meteorological Office. Brown, RE., 1997a. Adaptations ofcoral reefs to physical environmental stress. Advances in Marine Biology, 31: 221-299. Brown, B.E.• 1997b. Coral bleaching: causes and consequences. Coral Reefs. 16: SI29-S138. Brown. a.E., R.P. Dunne and H. Chansang. 1996. Coral bleaching relative to elevated seawater temperature in the Andaman Sea (Indian Ocean) over the last 50 years. Coral Reefs 15: 151­ 152. Folland, C.K. and D.E. Parker, 1995. Correction ofinstrumental biases in historical sea surface temperature data. Quart. J. Roy. Meteorol. Soc. 121: 319-367. Glynn, P.W., 1996. Coral reefbleaching: facts, hypotheses and implications. Global Change Biology, 2: 495-509. Houghton, J.T., L.G. Meira Filho, B.A. Callander, N. Harris, A. Kanenberg and K. Maskell (eds), 1996. Climate Change 1995. The Science ojClimate Change, Cambridge University Press. 572pp. Jones, RJ., R Berkelmans and J.K. Oliver, 1997. Recurrent bleaching ofcorals at Magnetic Island (Australia) relative to air and seawater temperature. Mar. &01. Prog. Ser.• 158: 289-292. Lough, J.M.• 1994. Climate variation and EI Niiio--Southern Oscillation events on the Great Barrier Reef. Coral Reefs 13: 181-195. Lough. J.M., 1997. Regional indices ofclimate variation: temperature and rainfall in Queensland, Australia. Int. J. Climatology 17: 55-66. Lough. J.M., 1998. Coastal climate ofnorthwest Australia and comparisons with the Great Barrier Reef: 1960 to 1992. Coral Reefs 17: 351-367. Mitchell, J.M., B. Dzerduevskii, H.Flohn, W.L. Hofmeyr, H.H. Lamb, K.N. Ran and C.C. Wallen, 1966. Climatic Change. World Meteorological Organi2ation Technical Note No. 79, WMO­ No. 195. TP 100, WMO, Geneva, 79pp. Pittock, A.B.. in press. Coral reefs and environmental change: adaptation to what? American Zoologist 39 Podesta, G.P. and P.W. Glynn, 1997. Sea surface temperature variability in Panama and Galapagos: extreme temperatures causing coral bleaching. J. Geaphys. Res. 102 (C7): 15749-15759. Reynolds, R W. and T.M. Smith, 1994. Improved global sea surface temperature analyses using optimum interpolation. J. Climate 7: 929.948. Ropelewski, C.F. and M.S. Halpert, 1996. Quantifying Sonthem Oscillation - precipitation relationships. J. Climate 9: 1043-1059. Wilkinson, C.R. (ed), 1998. Status ofCoral Reefs ofthe World: 1998. GCRMN, Australian Institute ofMarine Science. TownsviJIe. Australia, 184pp. Wilkinson, C.R and RW. Buddemeier, 1994. Global Climate Change and Coral Reefs: Implicationsfor People and Reefs. Report ofthe UNEP-IOC-ASPEI-!UCN Global Task Team on the implications ofclimate change on coral reefs. IUCN, Gland, Switzerland, 124pp. Wilkinson, C.R. O. Linden, H. Cesar, G. Hodgson, J. Rubens and A.E. Strong, in press. Ecological and socio-economic impacts of 1998 coral mortality in the Indian Ocean: an ENSO impact and a warning offuture change? Ambio. Woodruff, S.D., S.J. Lubker, K. Wolter, S.J. Worley and J.D. Elms, 1993. Comprehensive Ocean­ Atmosphere Data Set (COADS) Release la: 1980-1992. Earth System Monitor, 4: 1,4-8.

31